Improper device setting when using `neuron.device=cuda:<id>`

[ifrit98]

Using --neuron.device cuda:N fails with device placement mismatches.

e.g. When starting with
pm2 start run.sh --name openvalidators_autoupdate -- --wallet.name <your-wallet-name> --wallet.hotkey <your-wallet-hot-key> --neuron.device cuda:1

eventually fails with cuda:0 vs cuda:1 mismatch in forward() See: https://github.com/opentensor/validators/blob/main/openvalidators/forward.py#L297

Gets 1 tensor on cuda:0 and the other on cuda:1 set from the command line.

Suggestion is to simply recommend away from this arg until we find a fix.